Scoring Units on the Court Interpreter Oral Exam: What Raters Actually Count

The National Center for State Courts (NCSC) builds the oral exam used by most consortium states, including Florida, Texas and New Jersey. In NCSC's words, scoring units are "specific linguistic phenomena that interpreters must be able to deliver for a complete and accurate interpretation." Each unit is a word or short phrase the test writers chose on purpose because it tests something: a verb tense, a false cognate, a dollar amount, a curse word, a tag question tucked into a long sentence.

The candidate never sees the units. The rater's copy of the script has every one marked, and rating consists of deciding, unit by unit, whether your rendition preserved it. Everything else you said is evaluated only holistically. This is why two candidates who sounded similar can get very different scores: one lost the units and the other lost the filler.

One rule candidates skip past: when rating into Spanish, raters "will consider correct any rendering that would be acceptable in any appropriate spoken variety" of the language, and test writers avoid words peculiar to one country. A Mexican, Colombian or Puerto Rican equivalent is fine. An anglicism, a calque or a made-up word is not.

The NCSC Test Administration Standards (revised 2023) give fixed counts:

Do the arithmetic once and it changes how you practice. On the simultaneous, 70% of 75 is 52.5, so you can miss 22 units and pass; miss 23 and you fail. On the consecutive, you can miss 27 of 90. On sight, 70% of 50 is 35, but the 65% floor means you cannot rescue a bad Spanish-to-English passage (say 15 of 25) with a great English-to-Spanish passage. Some states set a stricter per-passage minimum; check your state's candidate overview.

One more rule with real consequences: in consecutive, if you run out of time, every unit after that point is counted wrong. In simultaneous, if you stop, the exam ends and every remaining unit is wrong. NCSC also scores only your last rendition if you give more than one, so a self-correction that lands is credited, and a self-correction that makes things worse costs you the unit.

NCSC's Test Construction Manual lists ten categories. The letters are NCSC's. The examples are ours, with Spanish renditions taken from state court glossaries so they match what raters accept (see the CourtFluent glossary for sources on each term).

• A. Grammar. Mostly verbs: tense, mood, person. "If I had known" rendered as "si supiera" instead of "si hubiera sabido" is a lost unit.
• B. Language interference. False cognates and calques. "Assault" is agresión, not asalto (which is robbery). "To apply for probation" is solicitar, not aplicar. "Eventually" is con el tiempo, not eventualmente.
• C. General vocabulary. Ordinary words a witness might use: "stumbled", "the trunk of the car", "a dent". This category tests range and the ability to produce a word "without hesitating or stumbling."
• D. Legal terms and phrases. "Bail" (fianza, also caución in Connecticut's glossary), "probation" (libertad a prueba, libertad probatoria), "parole" (libertad condicional, libertad bajo palabra), "bench warrant" (orden de arresto, orden judicial de arresto). NCSC says this category matters "especially from English into the other language."
• E. Idioms and sayings. "He was pulling my leg" needs me estaba tomando el pelo, not anything involving a leg.
• F. Register. The listener's impression of the speaker must not be "raised or lowered." A witness who says "yeah" did not say "yes", and a witness who swears must swear in Spanish.
• G. Numbers and names. Every digit, every letter of a spelled name in order, no unit conversion (miles stay miles), street and business names not translated.
• H. Markers, intensifiers, emphasis and precision. "Very", "only", "almost", "the day after tomorrow", "last night". Softening "he hit me hard" to me pegó loses the unit.
• I. Embeddings and position. Phrases that are easy to drop because of where they sit: the second adjective in a string, a mid-sentence aside, a tag question ("you did see him, didn't you?").
• J. Slang and colloquialisms. Meaning must be preserved; register does not have to be. "He was wasted" can be estaba borracho or andaba bien tomado; either works, but estaba desperdiciado is a miss.

Every exam is scored by two raters working independently. Each goes through the script and, per NCSC, places an X over any unit that is "omitted or rendered incorrectly." Then they compare. Where they disagree, they consult a scoring dictionary: a compilation of renditions past rater teams have accepted or rejected for that exact unit. If that doesn't settle it, they check bilingual dictionaries for any country where Spanish is spoken, and a third opinion can be sought. NCSC states the tiebreaker plainly: "The benefit of any doubt always goes to the candidate." If one rater believes the unit is acceptable after discussion, it is marked acceptable.

The score is subtraction. NCSC's example: 15 units marked unacceptable out of 75 leaves 60, and 60 divided by 75 is 80%. That number, per section, is what appears on your Results Report Form. What does not appear is why. NCSC is explicit that the form "does not include a holistic assessment or a diagnostic evaluation." You get a count, not a list.

There is one more layer. Raters also give a non-scoring-unit evaluation of English skills, Spanish skills and interpreting skills, rated Acceptable, Borderline or Unacceptable. Borderline does not affect the result. Unacceptable can: if both raters agree and a second rating team concurs, the candidate does not pass even at 70% on the units. It exists to catch a candidate who hits keywords while mangling the sentences around them, or who invents. It is rare, but it means the units are necessary, not sufficient.

Officially, a unit is either acceptable or unacceptable; NCSC does not award half points. But in practice there are renditions that would survive a scoring-dictionary discussion and renditions that would not, and it helps to see both. "Partial" below means a rendition that keeps the meaning but would make a rater reach for the dictionary; treat it as a coin flip and train it out.

Notice how many of the misses are not vocabulary gaps. The candidate knew what "bail" and "burglary" meant. They lost the unit to a digit, a dropped tag, a raised register or an anglicism produced under time pressure. That is what the exam is built to find, and it is why the consecutive and simultaneous guides spend so much time on numbers and lag rather than on word lists.

New Jersey is the only state that publishes section-level averages. For every Spanish candidate tested from November 1987 through December 2016 (2,630 people), the average percentage of units rendered correctly was:

The 2017 to 2023 data (184 Spanish candidates, 276 exams) shows higher averages, sight 68%, consecutive 65%, simultaneous 60%, but the same ordering and the same problem: only 39% of Spanish candidates reached 70% on simultaneous, versus 55% on sight. Simultaneous is where the count collapses, and the pass rate data across states says the same. The lowest score New Jersey ever recorded on each section is 0%, which is not a candidate who said nothing; it can be a candidate who stopped the simultaneous or ran out of consecutive time. The rules above are not fine print.

New Jersey's 2017 to 2023 numbers hide a detail: 184 Spanish candidates sat 276 exams, so a large share retested, and 36% of exams still ended in Fail with another 40% in Conditionally Approved (50% to 69%). Retakers improve on average, but many come back with roughly the same score, and the reason is structural. The Results Report Form gives you a count per section and nothing else. You don't learn which 23 units you lost on simultaneous, so you practice the things you noticed yourself: the term you blanked on, the sentence where you fell behind. The units you lost without noticing (a dropped "only", a raised register, a calque you use every day) stay lost, because nothing in your practice loop flags them.

The fix is to practice the way you are scored. Take a passage with marked units, record yourself, and mark your rendition unit by unit against the script, X for X. Count. Then sort your misses by category. Most candidates find that two or three of the ten categories account for most misses, and that those categories are almost never "legal terms." Numbers, embeddings, markers and interference are the usual suspects, and they are trainable in weeks once they are visible. The practice section walks through the mechanics for each mode. CourtFluent exists to make that loop cheap: it grades each recording against the scoring units, tags every miss by NCSC category, and turns misses into spaced-repetition flashcards. But the method works on paper too, if you are willing to be your own second rater and give yourself no benefit of the doubt.